Prism

dashboard
hl7 document processor
editor
patient list
triage

Inspiration

We watched clinical staff struggle with the cryptic nature of standard healthcare data formats like HL7, which is the standard for healthcare communication. Something as simple as checking a patient's most recent information often required physicians to call in IT support. That friction slows down admissions, discharges, and ultimately patient care. Prism was built to solve that problem. We bridge the gap between rigid technical standards and intuitive, human-friendly tools—so healthcare professionals can focus on treating patients, instead of deciphering code.

What it does

Prism is a clinician-friendly web app that translates HL7 v2 messages into clean, readable, and editable files. The platform features AI integrations that automatically extract and normalize clinical data from HL7 messages, validate medical information for consistency and completeness, and provide intelligent patient triage capabilities that analyze acuity levels and recommend priority rankings based on clinical indicators. Healthcare professionals can perform real-time editing of patient records, medications, and allergies through an intuitive interface, while the system generates comprehensive medical documents with proper formatting and exports updated records as professionally formatted PDFs, JSON for system integration, or regenerated HL7-compliant messages for seamless interoperability with existing hospital information systems. Prism also has a fast-acting triage mode that allows users to quickly assess which patients are in need of most urgent care.

How we built it

Our HL7 processing system is built on a FastAPI backend with a custom HL7 v2 parser for different message classes (admission, lab results, etc). Data flows through an AI pipeline orchestrated by Mastra.ai: Google's Gemini Pro, Claude, and ChatGPT work together to handle structured extraction, validation, and clinical reasoning for HL7 message processing and PDF generation. Parsed data is stored in a PostgreSQL database with proper relational models for patients, conversions, and audit logs.

The React/TypeScript frontend provides an intuitive dashboard built with TanStack Router, shadcn/ui components, and Framer Motion animations. Our Mastra service reconstructs compliant HL7 v2 messages and generates medical documents with configurable templates.

Challenges we ran into

The flexibility of the HL7 v2 standard was our biggest hurdle; parsing inconsistent message formats between different hospital systems required a highly configurable engine using LLMs. Ensuring data integrity after user edits was also critical. We implemented robust validation layers to guarantee all outbound messages remain syntactically correct.

Accomplishments that we're proud of

We’re most proud of building a tool that turns a notoriously complex problem into something simple and usable. To test it, we called our Program Manager at the Lundquist Institute (Harbor UCLA Medical Center) for a quick trial run. When it let him fire off multiple queries and receive clean, accurate results without friction, we couldn’t stop smiling. In that moment, we knew we had achieved our goal, a seamless translation layer that preserves full interoperability while dramatically improving usability (even if it has a few bugs.)